Determinants of Care Seeking for Children With Pneumonia and Diarrhea in Guatemala: Implications for Intervention Strategies

Abstract

Objectives. We identified barriers to care seeking for pneumonia and diarrhea among rural Guatemalan children.

Methods. A population-based survey was conducted twice from 2008 to 2009 among 1605 households with children younger than 5 years. A 14-day calendar recorded episodes of carer-reported pneumonia (n = 364) and diarrhea (n = 481), and formal (health services, public, private) and informal (neighbors, traditional, local shops, pharmacies) care seeking.

Results. Formal care was sought for nearly half of severe pneumonias but only for 27% within 2 days of onset, with 31% and 18%, respectively, for severe diarrhea. In multivariable analysis, factors independently associated with formal care seeking were knowing the Community Emergency Plan, mother’s perception of illness severity, recognition of World Health Organization danger signs, distance from the health center, and having someone to care for family in an emergency.

Conclusions. Proximal factors associated with recognizing need for care were important in determining formal care, and were strongly linked to social determinants. In addition to specific action by the health system with an enhanced community health worker role, a systems approach can help ensure barriers are addressed among poorer and more remote homes.

Globally, some 7.6 million children younger than 5 years die annually, the majority from low-income countries.¹ Two of the most important causes are acute lower respiratory infections, mainly pneumonia and diarrhea (18% and 14% of all deaths in children younger than 5 years, respectively). Adequate access to health care for young children with pneumonia and diarrhea is extremely important in improving survival as there are cheap and effective interventions available. In 2003, the Bellagio Child Survival Study group reviewed child survival interventions feasible for delivery in low-income settings, and concluded that if effective interventions had global coverage, 63% of childhood deaths could be prevented.² The power of existing interventions is not matched by the capacity of health systems to deliver them to those in greatest need in a comprehensive way and on an adequate scale.³ This situation persists, and a recent study of global research priorities for the prevention of deaths from pneumonia among children by 2015 identified barriers to care seeking and access as one of the highest priorities.⁴

In Guatemala, the 2006 infant mortality rate was 31 per 1000 live births, and under-5 mortality was 41 per 1000 child-years.⁵ In rural areas where the majority of the population resides, the infant mortality rate and under-5 mortality rates are likely considerably higher. On the basis of their investigation, Heuveline and Goldman suggested that improved access to health care could have considerable impact on under-5 mortality in Guatemala.⁶ Studies conducted in the 1990s found that health care uptake was very poor, with between 60% and 80% of families not seeking any formal, qualified health care for acute lower respiratory infections and diarrhea.^6,7 In relation to poor uptake, poverty was seen as an important predictor, and education and ethnicity were found to be less important.⁸

In addition to informal health care in Guatemala (traditional healers, midwives, neighbors, local stores, pharmacies), formal services are provided by the Ministry of Public Health and Social Assistance (MSPAS). The main health facilities include health posts (usually staffed by an auxiliary nurse), health centers (staffed by at least 1 doctor and qualified nurse), and “national” (general) and specialized hospitals. Recently there has been substantial growth in private services (individual doctors and hospitals), stimulated by rising income in urban areas and dissatisfaction with public services. Other health services are provided by the Social Security Institute and nongovernmental organizations. From 1997, MSPAS has also funded the Programme to Extend Coverage of Basic Health Services (Extension de Cobertura [EdC]). This program, part of the wider Comprehensive Health Care System (Sistema Integrado Atencion de Salud), involves the contracting of nongovernmental organizations to extend basic services to impoverished rural populations.⁹

Following a randomized controlled trial investigating the impact of reducing household air pollution on pneumonia among children,¹⁰ and motivated by efforts to develop integration of protection, prevention, and treatment,¹¹ we carried out a study in the same communities of Comitancillo and San Lorenzo into barriers to health care access. We used mixed methods (population-based surveys and qualitative interviews and focus groups) to understand the key demand and supply-side issues that could inform future interventions to improve access to high-quality care. We report here results from the survey component, and focus on the determinants of formal care seeking for maternal report of an episode in children of pneumonia, diarrhea, or both in the past 14 days, with “formal” care defined as contact with 1 or more MSPAS, private, or EdC services.

METHODS

The study communities were 2 municipalities within the rural western highlands of Guatemala in the Department (regional administrative area) of San Marcos. San Lorenzo is 35 kilometers from the town of San Marcos (the departmental main town, about 45 minutes by road), and Comitancillo is 30 kilometers farther north. Both areas are poor, relying mainly on subsistence farming. San Lorenzo, however, is somewhat more developed and affluent, with a higher proportion of Ladino people who are more acculturated to the Spanish-speaking culture. According to the 2006 survey, Encuesta Nacional de Condiciones de Vida: Guatemala, 35.5% of the San Marcos (departmental) population was indigenous, similar to the national figure of 38.4%, but San Marcos had considerably higher levels of extreme poverty at 21.1% compared with 15.2% nationally.¹² As shown in later tables, the proportion of indigenous population in both municipalities was considerably higher than for the whole of the San Marcos Department, and almost 100% in Comitancillo. There are 2 main seasons: the wet season occurring from mid-May until October, and the dry season, from October to May.

Sampling Methods and Survey

We sampled homes with children younger than 5 years from the 2 municipalities to allow comparison of care-seeking behavior and associated factors in communities known to differ in cultural, language, and other respects. In late 2007, we conducted a rapid census-type appraisal of 9503 homes, 6958 in Comitancillo and 2545 in San Lorenzo, excluding the main towns, to identify households with young children (< 5 years) and expectant mothers, from which families would be selected for the survey. For Comitancillo, which covered a large geographical area, we used stratified random sampling to select homes that were representative of one of the potential geographical barriers to health care access (distance from the health center), as follows. We listed communities (aldeas) by time on foot to the health center as estimated by the local health services, stratifying in 30-minute groups (range 30 minutes to > 4 hours), and labelled “large” or “small” according to whether the number of households was (1) at or above or (2) below the median for that stratum. We then calculated a target number of households from each stratum, proportional to the percentage of all homes in the study area provided by that stratum, and randomly sampled aldeas from the “high” and “low” groups within stratum until the target was reached (or exceeded), resulting in a total of 1462 homes. We included all 472 homes in San Lorenzo eligible at the time of the rapid appraisal, making a total for the study of 1934.

We used the survey to identify recent cases of community-acquired pneumonia and diarrhea and what care, if any, was sought and factors that influenced this. We carried out 2 rounds of structured interviews with mothers (or main carers) approximately 7 months apart collecting information by maternal or caregiver report on illness in the previous 14 days affecting the youngest child in the family. We conducted 2 sets of interviews to meet the sample size requirements—a logistic compromise because, to have sufficient households in a single interview, it would have meant carrying out the survey over a much wider area. This would have been very resource-intensive (time, vehicle usage, and costs) and difficult especially in the wet season when roads are often difficult or impossible to pass. We conducted the 2 surveys from October 2008 to May 2009, and June to December 2009, with informed consent obtained from all participants before the first interview. Interviews were carried out by local bilingual fieldworkers using either Spanish or Mam (the local Mayan language) according to interviewee preference. We conducted verbal autopsies on a sample of 30 deaths, which, in addition to obtaining the cause, was done to provide the basis for in-depth qualitative interviews on care sought and received, and which will be reported separately.

We conducted extensive development and pilot work including independent back-translation for all data collection instruments. A critical component was a 14-day calendar, which we used to record the illness and all associated care seeking and treatment, in the order in which this occurred, thus allowing “pathways” through multiple informal and formal care providers to be identified. Although based on earlier use of such calendars,¹³ we developed the instrument used in the current study specifically for this purpose, and tested it over a period of 8 weeks to allow the field team to become familiar with its use through close supervision and feedback by research staff. We carried out some validation of this methodology through comparison with assessments made by nurse auxiliaries of a representative sample of pneumonia and diarrhea cases. The nurse auxiliary visits were carried out on the same day as the survey to (1) ensure appropriate advice and help, and (2) provide independent validation of field worker assessment. The nurse auxiliary repeated the standard questions, and carried out an examination of clinical signs.

Illness and Care-Seeking Definitions

On the basis of responses during interview, we classified cases of “severe pneumonia” and “severe diarrhea” according to the reported presence of key symptoms and signs used for the World Health Organization (WHO) definition of community-acquired pneumonia and diarrhea.^14,15 The primary outcome for this study was, among children with severe pneumonia (requiring presence of lower chest wall indrawing or general danger signs) or severe diarrhea (requiring presence of sunken eyes or general danger signs) or both, care sought from formal health services. This outcome included (1) care sought at any time during the illness and (2) care sought within 2 days of onset. This latter definition served as a measure of prompt care seeking.

We investigated a range of characteristics in relation to care seeking, including child, mother and household factors. For the child, these were season of illness, age, sex, and ethnicity. Field staff measured height and weight at clinic sessions organized for the study, and used stunting (height > 2 SDs below median of WHO Child Growth Standards) and underweight (weight > 2 SDs below norm) in analysis.¹⁶ For the mother, these were number of children in the family, experience of previous child death, literacy, educational level, perception of illness severity, knowledge of Community Emergency Plan (CEP),¹⁷ recognition of WHO danger signs for acute lower respiratory infections and diarrhea, and belief in mal de ojo (“evil eye”) as a cause of illness. For the household, these were municipality, distance from main town, altitude, main language spoken, male employment status, crowding, access to own transport, access to phone, asset index score (ownership of radio, television, or refrigerator, each scoring 1, and added together giving a range of 0–3), and someone available to look after household in emergency.

The CEP was originally developed by the US Agency for International Development, administered by the Guatemalan MSPAS, and targeted at families in more remote areas.¹⁷ It provides picture-based guidance in seeking prompt, appropriate care for childhood and adult illness. Four of the WHO danger signs were common to severe pneumonia and severe diarrhea (general danger signs): the child “being unable to suckle or drink,” “vomiting everything,” “having convulsions,” or “being unconscious or lethargic.”¹⁴ One additional sign of severe pneumonia was “lower chest wall indrawing,” and for severe diarrhea, additional signs were “blood in stools” and “sunken eyes.” We assessed knowledge of the plan at interview by asking the parents if they were familiar with it, while showing them a copy. Thus, this did not test knowledge of everything covered by the CEP, but this is in part covered by the questions on knowledge of the WHO danger ;signs.

Statistical Analysis

We used unconditional logistic regression to explore independent relationships between child, mother, and household characteristics with care-seeking outcomes, with 4 separate models covering any formal care seeking and care seeking within 2 days, for the 2 health outcomes. We set an entry criterion of P < .2 in univariable analysis for inclusion in multivariable regression.

We used a multilevel (random intercepts) framework employing the GLLAMM procedure to adjust for children represented more than once in the analysis (i.e., a case in both survey rounds), and which provides robust standard errors. We analyzed data with SPSS for Windows version 18.0 (SPSS Inc, Chicago, IL) and Stata version 10 (StataCorp LP, College Station, TX). All P values are 2-sided.

RESULTS

We conducted a total of 3151 interviews, 1605 for survey 1 (83% response rate) and 1546 (96.3% re-interviewed) for survey 2. For survey 1, 403 (25.1%) resided in San Lorenzo and 1202 (74.9%) in Comitancillo. The majority of children were from the indigenous Mam linguistic group (93.2%), were stunted (60.8%), and were aged between 1 and 5 years (74.7%; Table A, available as a supplement to the online version of this article at http://www.ajph.org). We identified significant differences between the 2 municipalities: children from Comitancillo were more likely to be female, of indigenous ethnicity, stunted, and underweight, and were less likely to have had solid food introduced by 6 months (P < .05).

The majority of mothers were aged 25 years or older (68%), were literate (65.5%), had some formal education (71.7%), and did not report previous death of a child (91.9%). More than 40% had 3 or more children in addition to the index child. Mothers in Comitancillo were generally older, had larger families, and were more likely to be illiterate, to be without formal education, and to have experienced a previous death of a child (P < .05).

The majority of households were located more than 5 kilometers from the main town (57.5%), included family members who spoke Spanish (60.8%), and had the male head of the family working as a laborer (someone paid on a daily rate for working on another’s land, in a workshop, etc.; 58.2%). More than half used a 3-stone open fire for cooking (55%) with only 6% having access to their own car and owning a refrigerator. Households in Comitancillo were significantly more likely to be more than 5 kilometers from the main town, to speak Mam in the household, to have a male head of the household working as a laborer, to be overcrowded, to have only an open fire for cooking, and to have no ownership of a car, phone, radio, TV, or refrigerator (P < .05).

Although reported knowledge of the CEP may not be wholly reliable as a measure of actual familiarity with details of the plan and readiness to act on the advice, it does provide some indication of this. Knowledge of the CEP was significantly lower in Comitancillo, among indigenous households, in those speaking Mam, for mothers who were illiterate and less educated, and among farming households (Table 1). Knowledge was also lower among those with no phone, living farther from the main town, and where there had been a previous child death. Mothers reporting that there was no one else to look after the family in an emergency were more likely to know the CEP. In multivariable analysis, significant associations remained for a number of these variables, but not for municipality and language spoken at home (Table 1). Perhaps surprisingly, previous death of a child and having someone to look after the family in an emergency were associated with a lower likelihood of knowing the CEP.

TABLE 1—

Associations Between Socioeconomic Factors and Knowledge of the Community Emergency Plan: San Marcos, Guatemala, 2008–2009

Characteristic	No. of Mothers Knowing CEP (%)	OR (95% CI)	P	AOR (95% CI)	P
Municipality
Comitancillo	664 (55.2)	1.00 (Ref)		1.00 (Ref)
San Lorenzo	282 (70.1)	1.90 (1.50, 2.43)	<.001	1.39 (0.98, 1.97)	.07
Ethnicity
Ladino	83 (76.9)	1.00 (Ref)		1.00 (Ref)
Indigenous	862 (57.8)	0.43 (0.27, 0.67)	<.001	0.78 (0.50, 1.33)	.36
Language at home
Mam only (Ref)	681 (55.5)	1.00 (Ref)		1.00 (Ref)
Spanish	176 (72.1)	2.07 (1.53, 2.80)	<.001	1.26 (0.98, 1.64)	.07
Mother literate
Yes	665 (63.4)	1.00 (Ref)		1.00 (Ref)
No	176 (50.5)	0.80 (0.73, 0.87)	<.001	0.83 (0.48, 1.49)	.57
Mother’s education
None	228 (50.2)	1.00 (Ref)		1.00 (Ref)
Elementary	537 (62.6)	1.66 (1.32, 2.09)	<.001	1.28 (0.70, 2.34)	.43
> elementary	181 (62.0)	1.62 (1.19, 2.18)	.002	1.33 (0.82, 2.17)	.25
Previous death of child
No	889 (60.3)	1.00 (Ref)		1.00 (Ref)
Yes	57 (43.8)	0.51 (0.36, 0.74)	<.001	0.57 (0.38, 0.86)	.007
Possess phone
Yes	735 (61.1)	1.00 (Ref)		1.00 (Ref)
No	211 (52.6)	0.71 (0.56, 0.88)	.003	0.84 (0.64, 1.10)	.2
Overcrowding
No	499 (60.7)	1.00 (Ref)		1.00 (Ref)
Yes	447 (57.2)	0.86 (0.71, 1.03)	.15	1.02 (0.81, 1.30)	.81
Someone to look after family in emergency
No	258 (64.5)	1.00 (Ref)		1.00 (Ref)
Yes	688 (57.1)	0.88 (0.81, 0.97)	.01	0.63 (0.48, 0.82)	.001
Employment status of head of household
Farmer (Ref)	102 (48.6)	1.00 (Ref)		1.00 (Ref)
Laborer	554 (60.5)	1.62 (1.20, 2.19)	.002	2.13 (1.50, 3.02)	.001
Self-employed	77 (59.7)	1.57 (1.01, 2.44)	.04	1.30 (0.79, 2.14)	.3
Other	192 (60.4)	1.61 (1.14, 2.29)	.008	1.67 (1.11, 2.52)	.01
Access to transport
None	685 (58.7)	1.00 (Ref)
Bike or motorcycle	193 (58.5)	0.99 (0.77, 1.27)	.95
Car or pickup truck	68 (63.6)	1.22 (0.81, 1.85)	.33
Asset indexa
Score 0–1	539 (55.6)	1.00 (Ref)		1.00 (Ref)
Score 2–3	407 (64.1)	1.42 (1.15, 1.75)	.001	1.39 (1.08, 1.79)	.01
Distance from town: 1 km closer		1.06 (1.03, 1.08)	<.001	01.02 (0.99, 1.06)	.16

Note. AOR = adjusted odds ratio; CEP = Community Emergency Plan; CI = confidence interval; OR = odds ratio. Mam is the local Mayan language.

^aBased on radio, TV, or refrigerator ownership.

We found somewhat weaker associations between knowing 1 or more WHO danger signs for pneumonia and diarrhea and municipality, ethnicity, and having someone to look after the family in an emergency (pneumonia only; Table 2). In multivariable analysis, only asset index remained significantly (P = .03) associated with knowledge of danger signs for diarrhea, albeit negatively.

TABLE 2—

Associations Between Socioeconomic Factors and Knowledge of the World Health Organization Danger Signs for Severe Pneumonia and Severe Diarrhea: San Marcos, Guatemala, 2008–2009

Characteristic	No. of Mothers Knowing WHO Danger Signsa (%)	OR (95% CI)	P	AOR (95% CI)	P
Severe pneumonia
Municipality
Comitancillo	353 (29.4)	1.00 (Ref)		1.00 (Ref)
San Lorenzo	150 (37.2)	1.42 (1.13, 1.81)	.003	1.32 (0.95, 1.86)	.1
Ethnicity
Ladino	44 (40.4)	1.00 (Ref)		1.00 (Ref)
Indigenous	458 (30.7)	0.66 (0.44, 0.98)	.04	0.94 (0.58, 1.52)	.81
Language at home
Mam only	367 (29.9)	1.00 (Ref)		1.0
Spanish	87 (35.7)	1.29 (0.97, 1.73)	.08	0.93 (0.72, 1.21)	.6
Mother literate
Yes	332 (31.6)	1.00 (Ref)
No	171 (30.9)	0.98 (0.84, 1.14)	.76
Mother’s education
None	134 (29.5)	1.00 (Ref)
Elementary	266 (31.0)	1.07 (0.84, 1.38)	.58
> elementary	103 (35.2)	1.30 (0.95, 1.77)	.11
Previous death of child
No	466 (31.6)	1.00 (Ref)
Yes	37 (28.5)	0.86 (0.58, 1.28)	.46
Possess phone
Yes	392 (32.6)	1.00 (Ref)		1.00 (Ref)
No	111 (27.7)	0.79 (0.62, 1.02)	.07	0.83 (0.63, 1.10)	.2
Overcrowding
No	265 (32.2)	1.00 (Ref)
Yes	238 (30.4)	0.92 (0.75, 1.14)	.45
Someone to look after family in emergency
No	108 (27.0)	1.00 (Ref)		1.00 (Ref)
Yes	395 (32.8)	1.21 (1.01, 1.45)	.03	0.79 (0.58, 1.03)	.09
Employment status of head of household
Farmer	73 (34.8)	1.00 (Ref)
Laborer	276 (30.1)	0.81 (0.59, 1.11)	.19
Self-employed	41 (31.8)	0.87 (0.55, 1.40)	.57
Other	106 (33.9)	0.94 (0.64, 1.36)	.73
Access to transport
None	366 (31.4)	1.00 (Ref)
Bike or mortorcycle	100 (30.3)	0.95 (0.73, 1.24)	.71
Car or pickup truck	37 (34.3)	1.14 (0.84, 1.30)	.54
Asset indexb
Score 0–1	300 (31.0)	1.00 (Ref)
Score 2–3	203 (31.9)	1.04 (0.84, 1.30)	.69
Distance from town: 1 km closer		1.02 (0.99, 1.05)	.07	1.02 (0.98, 1.05)	.32
Severe diarrhea
Municipality
Comitancillo	367 (30.5)	1.00 (Ref)		1.00 (Ref)
San Lorenzo	144 (35.7)	1.27 (0.99, 1.61)	.05	1.09 (0.80, 1.48)	.58
Ethnicity
Ladino	45 (40.9)	1.00 (Ref)		1.00 (Ref)
Indigenous	465 (31.2)	0.65 (0.44, 0.97)	.04	0.74 (0.47, 1.16)	.19
Language at home
Mam only	385 (31.4)	1.00 (Ref)
Spanish	80 (32.8)	1.07 (0.79, 1.43)	.67
Mother literate
Yes	341 (32.5)	1.00 (Ref)
No	169 (30.5)	1.10 (0.88, 1.37)	.42
Mother’s education
None	137 (30.2)	1.00 (Ref)
Elementary	274 (31.9)	1.09 (0.85, 1.39)	.51
> elementary	100 (34.1)	1.20 (0.88, 1.64)	.26
Previous death of child
No	465 (31.5)	1.00 (Ref)
Yes	46 (35.4)	1.18 (0.82, 1.73)	.37
Possess phone
Yes	394 (32.7)	1.00 (Ref)		1.00 (Ref)
No	117 (29.2)	0.85 (0.66, 1.08)	.19	0.81 (0.62, 1.06)	.12
Overcrowding
No	268 (32.6)	1.00 (Ref)
Yes	243 (31.1)	0.93 (0.76, 1.15)	.52
Someone to look after family in emergency
No	120 (30.0)	1.00 (Ref)
Yes	391 (32.4)	1.08 (0.91, 1.28)	.36
Employment status of head of household
Farmer	69 (32.9)	1.00 (Ref)
Laborer	284 (31.0)	0.92 (0.67, 1.27)	.6
Self-employed	41 (31.8)	0.95(0.16, 1.52)	.84
Other	111 (34.9)	1.10 (0.76, 1.59)	.63
Access to transport
None	380 (32.6)	1.00 (Ref)
Bike or motorcycle	96 (29.1)	0.85 (0.65, 1.11)	.23
Car or pickup truck	35 (32.4)	0.99 (0.65, 1.51)	.97
Asset indexb
Score 0–1	323 (33.3)	1.00 (Ref)		1.00 (Ref)
Score 2–3	188 (29.6)	0.84 (0.67, 1.04)	.11	0.70 (0.50, 0.96)	.03
Distance from town: 1 km closer		1.03 (0.99, 1.06)	.06	1.02 (0.99, 1.06)	.17

Note. AOR = adjusted odds ratio; CI = confidence interval; OR = odds ratio; WHO = World Health Organization. Mam is the local Mayan language.

^aIncludes the WHO general danger signs and those specific to the disease.

^bBased on radio, TV, or refrigerator ownership.

Based on responses to the 3151 interviews relating to recent (previous 14 days) health of children, the number (prevalence) of cases of severe pneumonia and severe diarrhea were 364 (11.6%) and 481 (15.3%), respectively. There were 185 cases of severe pneumonia in round 1 and 179 in round 2, with 31 having an episode in both surveys. For severe diarrhea, there were 232 cases in round 1 and 249 in round 2, with 59 having an episode in both surveys. In the nurse auxiliary validation study, which analyzed 272 cases of possible community-acquired pneumonia according to maternal report that were also examined by the nurse auxiliary, a total of 147 (54%) were confirmed by nurse auxiliary assessment (κ = 0.49). For the 472 survey-defined cases of diarrhea, 396 (83.9%) were confirmed by nurse auxiliary (κ = 0.74). Analysis of factors associated with care seeking among nurse auxiliary ascertained cases generally confirmed results from the main survey.

First Source of Care Sought

Among the 364 severe pneumonia cases, around 5% of families sought no care, from either formal or informal health care providers (Table 3). For a little more than 20%, the first point of contact for advice was one of the formal health care providers, but informal agencies were more frequently the first contact point (local store, 24.7%; relative, 18.4%; pharmacy, 16.5%). In San Lorenzo, 39% of mothers with a child with severe pneumonia went to a health center or health post first compared to only 14.1% in Comitancillo (P < .001), although the local store was commonly approached first in both municipalities (25.5% and 22.9%, respectively; P = .61).

TABLE 3—

First Agency Contacted for Health Care and Advice Relating to the Sick Child Younger Than 5 Years: San Marcos, Guatemala, 2008–2009

	Severe Pneumonia Episodes, %			Severe Diarrhea Episodes, %
Agency First Sought for Help	All (n = 364)	Comitancillo (n = 255)	San Lorenzo (n = 109)	All (n = 481)	Comitancillo (n = 371)	San Lorenzo (n = 110)
No help sought	4.7	5.9	1.8	6.7	5.5	10.9
Community providers	65.2	68.7	57.7	77.6	81.1	65.6
Relative	18.4	20.8	12.8	21.0	21.3	20.2
Neighbors	3.8	3.9	3.7	6.9	7.8	3.6
Healer	0.8	0.8	0.9	1.5	1.6	0.9
Midwife	0.5	0.8	0	0.6	0.8	0
Medicine seller	0.5	0.4	0.9	1.2	1.3	0.9
Local store	24.7	25.5	22.9	35.8	37.5	30.0
Pharmacy	16.5	16.5	16.5	10.6	10.8	10.0
MSPAS and EdC	26.1	20.4	36.7	13.1	10.8	20.9
EdC facilitator	5.2	5.5	0	3.3	4.3	0
Health post	6.0	4.3	10.1	2.1	0.8	6.4
Health center	14.1	9.8	25.7	7.7	5.7	14.5
National hospital	0.8	0.8	0.9	0	0	0
Private	3.3	3.4	3.7	2.5	2.4	2.7
Private doctor	1.9	1.6	2.8	1.0	0.8	1.8
Private hospital	1.4	1.6	0.9	1.5	1.6	0.9

Note. EdC = Programme to Extend Coverage of Basic Health Services (Extension de Cobertura); MSPAS = Ministry of Public Health and Social Assistance. Chi-squared test for comparison between municipalities at level of headings (no help sought, community providers, MSPAS and EdC, and private): Comitancillo severe pneumonia episodes, P = .03; all severe diarrhea episodes, P = .009.

For the 481 severe diarrhea cases, 7% of mothers sought no care at all. The local store was used more frequently than for severe pneumonia (36% compared with 25% for severe pneumonia). A higher proportion of severe diarrhea cases in San Lorenzo were initially taken to a health center or post than in Comitancillo (21% vs 6.5%, respectively; P < .001).

Formal Care for Severe Pneumonia

Of the 364 cases of severe pneumonia, formal care was sought for 181 (49.7%). For a quarter (n = 98) of cases, formal care was sought within 2 days of onset. Table 4 shows associations between child, mother, and household characteristics and care-seeking outcomes for which the entry criterion of P < .2 in univariable analysis was met. Factors significantly associated in univariable analysis with formal care seeking at any time in the episode (P < .05) included a higher level of maternal education, perception of a greater illness severity, knowledge of the CEP, San Lorenzo residence, and recognition of WHO danger signs. Multivariable analysis identified 2 independent factors for formal care seeking: mother’s perception of severity (adjusted odds ratio [OR] = 2.05; 95% confidence interval [CI] = 1.05, 4.03) and maternal recognition of WHO danger signs (adjusted OR = 2.14; 95% CI = 1.32, 3.47).

TABLE 4—

Associations for Factors Related to Formal Care Seeking for Children Younger Than 5 Years With Severe Pneumonia and Severe Diarrhea: San Marcos, Guatemala, 2008–2009

Variable	No. of Episodes (%)	Unadjusted OR (95% CI)	P	Adjusted ORa (95% CI)	P
Severe pneumonia: formal care seeking any time during episode
Age of child
< 12 mo	66 (55.5)	1.00 (Ref)		1.00 (Ref)
1–5 y	114 (46.9)	0.71 (0.46, 1.10)	.13	0.73 (0.44, 1.21)	.22
Gender of child
Male	97 (55.4)	1.00 (Ref)		1.00 (Ref)
Female	83 (45.6)	0.67 (0.44, 1.02)	.06	0.76 (0.48, 1.21)	.25
Underweight (weight 2 SD below norm)
No	123 (53.0)	1.00 (Ref)		1.00 (Ref)
Yes	49 (44.5)	0.71 (0.45, 1.12)	.14	0.83 (0.50, 1.38)	.47
Mother’s education level
None	43 (44.8)	1.00 (Ref)		1.00 (Ref)
Elementary	98 (48.5)	1.16 (0.71, 1.89)	.55	0.97 (0.56, 1.68)	.91
> elementary	40 (60.6)	1.90 (1.01, 3.59)	.04	1.53 (0.74, 3.17)	.25
Mother’s perceived severity of illness episode
Not serious	24 (42.1)	1.00 (Ref)		1.00 (Ref)
Serious	77 (45.6)	1.15 (0.63, 2.11)	.65	1.11 (0.58, 2.14)	.75
Very serious	80 (58.4)	1.93 (1.03, 3.61)	.04	2.05 (1.05, 4.03)	.04
Mother knows emergency plan
No	40 (39.2)	1.00 (Ref)		1.00 (Ref)
Yes	141 (53.8)	1.81 (1.13, 2.88)	.01	1.59 (0.97, 2.62)	.07
Municipality
Comitancillo	117 (45.9)	1.00 (Ref)		1.00 (Ref)
San Lorenzo	64 (58.7)	1.67 (1.06, 2.64)	.03	1.65 (0.96, 2.84)	.07
Asset index scoreb
0–1	105 (46.9)	1.00 (Ref)		1.00 (Ref)
2–3	76 (54.3)	1.35 (0.88, 2.06)	.17	1.29 (0.79, 2.09)	.31
Mother recognized WHO sign
No	52 (37.7)	1.00 (Ref)		1.00 (Ref)
Yes	129 (57.1)	2.20 (1.43, 3.39)	.001	2.14 (1.32, 3.47)	.002
Severe pneumonia: formal care seeking within 2 days of onset
Gender of child
Male	54 (30.9)	1.00 (Ref)		1.00 (Ref)
Female	43 (23.6)	0.69 (0.43, 1.11)	.13	0.83 (0.50, 1.35)	.44
Mother experienced previous death of a child
No	93 (28.1)	1.00 (Ref)		1.00 (Ref)
Yes	5 (15.2)	0.46 (0.17, 1.22)	.11	0.49 (0.18, 1.34)	.17
Mother’s literacy
Illiterate	21 (18.9)	1.00 (Ref)		1.00 (Ref)
Literate	77 (30.4)	1.88 (1.09, 3.24)	.02	2.13 (0.60, 7.58)	.24
Mother’s education level
None	18 (18.8)	1.00 (Ref)		1.00 (Ref)
Elementary	58 (28.7)	1.75 (0.98, 3.17)	.07	0.73 (0.18, 2.94)	.66
> elementary	22 (33.3)	2.17 (1.05, 4.47)	.04	0.97 (0.27, 3.51)	.96
Mother knows emergency plan
No	21 (20.6)	1.00 (Ref)		1.00 (Ref)
Yes	77 (29.4)	1.61 (0.93, 2.78)	.09	1.46 (0.83, 2.56)	.19
Municipality
Comitancillo	62 (24.3)	1.00 (Ref)		1.00 (Ref)
San Lorenzo	36 (33.0)	1.53 (0.94, 2.51)	.09	1.11 (0.65, 1.89)	.70
Someone to look after household in emergency
No	24 (21.2)	1.00 (Ref)		1.00 (Ref)
Yes	74 (29.5)	1.55 (0.92, 2.62)	.10	1.43 (0.83, 2.49)	.20
Location of household: 1 km closer to town		1.06 (0.99, 1.15)	.05	1.08 (0.99, 1.16)	.06
Severe diarrhea: formal care seeking any time during episode
Age of child
< 12 mo	47 (36.7)	1.00 (Ref)		1.00 (Ref)
1–5 y	100 (28.6)	0.69 (0.45, 1.06)	.09	0.79 (0.49, 1.26)	.32
Ethnic group of child
Non-Indigenous	12 (52.2)	1.00 (Ref)		1.00 (Ref)
Indigenous	135 (29.7)	0.39 (0.17, 0.90)	.02	0.55 (0.20, 1.49)	.24
Mother’s literacy
Illiterate	45 (26.2)	1.00 (Ref)		1.00 (Ref)
Literate	102 (33.1)	1.39 (0.92, 2.12)	.11	1.23 (0.48, 3.19)	.67
Mother’s education level
None	36 (24.8)	1.00 (Ref)		1.00 (Ref)
Elementary	76 (31.3)	1.38 (0.87, 2.19)	.18	0.97 (0.33, 2.81)	.95
> elementary	35 (38.0)	1.86 (1.06, 3.27)	.03	1.31 (0.52, 3.34)	.57
Mother’s perceived severity of illness episode
Not serious	16 (21.6)	1.00 (Ref)		1.00 (Ref)
Serious	69 (27.5)	1.37 (0.74, 2.55)	.31	1.23 (0.66, 2.40)	.53
Very serious	62 (40.0)	2.42 (1.27, 4.58)	.007	2.22 (1.14, 4.35)	.02
Mother knows emergency plan
No	38 (24.1)	1.00 (Ref)		1.00 (Ref)
Yes	109 (33.9)	1.62 (1.05, 2.49)	.03	1.73 (1.09, 2.76)	.02
Municipality
Comitancillo	105 (28.4)	1.00 (Ref)		1.00 (Ref)
San Lorenzo	42 (38.2)	1.56 (0.99, 2.44)	.052	1.21 (0.66, 2.22)	.55
Mother believed in mal de ojo
No	138 (31.9)	1.00 (Ref)		1.00 (Ref)
Yes	9 (19.1)	0.51 (0.24, 1.08)	.07	0.58 (0.27, 1.27)	.18
Location of household: 1 km closer to town		1.11 (1.05, 1.18)	.001	1.11 (1.04, 1.18)	.002
Severe diarrhea: formal care seeking within 2 days of onset
Ethnic group of child
Non-Indigenous	8 (34.8)	1.00 (Ref)		1.00 (Ref)
Indigenous	78 (17.1)	0.39 (0.16, 0.95)	.03	0.63 (0.22, 1.81)	.39
Mother experienced previous death of a child
No	81 (18.7)	1.00 (Ref)		1.00 (Ref)
Yes	5 (10.6)	0.52 (0.20, 1.35)	.17	0.69 (0.25, 1.88)	.47
Mother’s literacy
Illiterate	24 (14.0)	1.00 (Ref)		1.00 (Ref)
Literate	62 (20.1)	1.55 (0.93, 2.60)	.09	1.31 (0.74, 2.34)	.36
Mother’s perceived severity of illness episode
Not serious	9 (12.2)	1.00 (Ref)		1.00 (Ref)
Serious	43 (17.1)	1.49 (0.69, 3.23)	.31	1.32 (0.59, 2.95)	.51
Very serious	34 (21.9)	2.03 (0.92, 4.49)	.08	1.66 (0.72, 3.87)	.24
Mother knows emergency plan
No	23 (14.6)	1.00 (Ref)		1.00 (Ref)
Yes	63 (19.6)	1.43 (0.85, 2.40)	.18	1.58 (0.88, 2.80)	.12
Municipality
Comitancillo	57 (15.4)	1.00 (Ref)		1.00 (Ref)
San Lorenzo	29 (26.4)	1.97 (1.18, 3.27)	.009	1.32 (0.62, 2.81)	.47
Main language spoken in household
Mam only	28 (15.6)	1.00 (Ref)		1.00 (Ref)
Spanish first or second language	54 (20.5)	1.40 (0.85, 2.32)	.19	0.85 (0.46, 1.56)	.59
Someone to look after household in emergency
No	22 (14.4)	1.00 (Ref)		1.00 (Ref)
Yes	64 (19.6)	1.45 (0.86, 2.46)	.17	2.01 (1.10, 3.68)	.02
Mother believed in mal de ojo
No	82 (18.9)	1.00 (Ref)		1.00 (Ref)
Yes	4 (8.5)	0.40 (0.14, 1.14)	.08	0.44 (0.15, 1.32)	.14
Location of household: 1 km closer to town		1.14 (1.05, 1.22)	.001	1.12 (1.03, 1.22)	.007

Note. CI = confidence interval; mal de ojo = "evil eye"; OR = odds ratio; WHO = World Health Organization. Formal care includes all government and private medical services including extension of cover (Programme to Extend Coverage of Basic Health Services [Extension de Cobertura]). Mam is the local Mayan language.

^aAdjusted for all variables listed in unadjusted column (P < .02 in univariable analysis).

^bBased on radio, TV, or refrigerator ownership.

In relation to prompt (within 2 days) formal care seeking, 2 factors were significantly (P < .05) associated in univariable analysis: maternal literacy and higher level of maternal education. None of the household, child, and mother characteristics were significantly associated with prompt care seeking in multivariable analysis, although distance to town was of borderline significance (1 km closer OR = 1.08; 95% CI = 0.99, 1.16). Analysis of factors independently associated with formal care seeking for severe pneumonia within 2 days among the subset of 181 cases seeking formal care at any time did not identify any significant findings.

Formal Care for Severe Diarrhea

Of the 480 cases of severe diarrhea, formal help was sought for 147 (30.6%). For almost a fifth (n = 86), formal help was sought within 2 days. For formal care seeking at any time during the episode (Table 4), we observed a significant association in univariable analysis (P < .05) for a higher level of maternal education, greater perceived severity of illness, knowledge of CEP, child’s ethnic group being Ladino, and a closer proximity to the main town. Three factors were independently associated with formal care seeking in the multivariable model: mother’s perception of severity (adjusted OR = 2.22; 95% CI = 1.14, 4.35), knowledge of the CEP (adjusted OR = 1.73; 95% CI = 1.09, 2.76), and distance from the main town (1 km closer, adjusted OR = 1.11; 95% CI = 1.04, 1.18). Three factors were significantly (P < .05) associated in univariable analysis with prompt formal care seeking (Table 4): Ladino ethnicity, San Lorenzo residence, and proximity to the main town.

We found 2 factors to be independently associated with prompt care seeking in the multivariable model: proximity to the main town (1 km closer, adjusted OR = 1.12; 95% CI = 1.03, 1.22) and availability of someone to look after the household in an emergency (adjusted OR = 2.01; 95% CI = 1.10, 3.68). Analysis of factors independently associated with formal care seeking for severe diarrhea within 2 days among the subset of 147 cases seeking formal care at any time identified someone to look after the children in an emergency (OR = 2.98; 95% CI = 1.48, 6.62) as the only significant finding.

DISCUSSION

In this study, recent and current episodes of pneumonia and diarrhea were identified by parental interview using recall of established criteria for community cases. The reported prevalence rates for severe pneumonia (11.6%) and severe diarrhea (15.3%) are high, but consistent with estimates derived from similar questions used, for example, in Demographic and Health Surveys,¹⁸ and reflect the sensitive but nonspecific characteristics of this method.¹⁹ The nurse auxiliary validation study showed reasonable agreement with fieldworker interviews, allowing for the expected lower specificity of the latter. Furthermore, it is important that care is sought for sick children who—on the basis of symptoms or signs apparent to parents—may have severe pneumonia or diarrhea (and not just clinically confirmed cases). Therefore, despite the sensitive but nonspecific case ascertainment, the findings are relevant to the primary goal of this work—to inform interventions for improving more complete and rapid care seeking.

Formal and Informal Sources of Care

Sick children may deteriorate rapidly, so time taken to obtain trained medical attention is very important, whatever the initial source of advice. In our study, the most common source of initial help was informal, with children from Comitancillo being less likely to first attend a health center or health post. The local store and pharmacy were popular as sources of initial advice, regardless of municipality or the specific child’s illness.

Pharmacies may provide some degree of formal care, either in assessment of a sick child, or in the direct or indirect (following medical assessment) prescription of appropriate treatment. We found that 63 and 70 cases of severe pneumonia and severe diarrhea, respectively, involved a pharmacy visit (either initial or subsequent), but in only 2 and 1 of these, respectively, was this subsequent to a formal (MSPAS or EdC) visit. The pharmacy was the source of antibiotics in 22% (n = 18) of severe pneumonia cases that received these, and of oral rehydration solutions in 12% (n = 9) of the severe diarrhea cases that received these, so most of these must have been direct over-the-counter dispensing. Thus, a minority of pharmacy visits may have provided something akin to formal care, but full assessment of the appropriateness of this would require further investigation. Among the more important findings from the qualitative component of the study were that mothers placed much faith in herbal remedies, treatments known to be unavailable and disapproved of in the MSPAS facilities, but easily available in the local store or pharmacy.

Factors Associated With Formal Care Seeking

Previous research from Africa, South Asia, and Latin America of factors associated with decreased likelihood to seek formal care for sick children has identified lower educational level,^20,21 younger maternal age,²² indigenous ethnicity,^23,24 lower occupational status,⁷ lack of perceived illness severity,^8,25 and death of a previous child.²⁶ In addition, child’s age,²⁷ nutritional status,²⁸ and socioeconomic factors^20,29–34 have been reported as being associated with care-seeking behavior. Although indigenous ethnicity might be thought to be a useful indicator and basis for analysis, our understanding (from observation and local partners) is that those describing themselves as indigenous in the more Ladino, acculturated San Lorenzo differ from those in Comitancillo. This is borne out by the finding that 95.9% of indigenous families in San Lorenzo speak Spanish at home compared with only 48.7% in Comitancillo.

In our survey, many of these factors (or closely associated variables) were associated with formal care seeking in univariable analysis, but were not independent predictors in multivariable analysis. The factors that stood out as independent predictors were the mother’s perception of the child’s illness severity, knowledge of the emergency plan, and the ability to identify WHO signs of severe illness. Distance from the main town was also significant for diarrhea (P = .002; formal care at any time), but not for formal care for pneumonia sought within 2 days (P = .06). These findings lend support to the potential effectiveness of comprehensive educational interventions, focusing on prompt recognition of warning signs and preparing for an emergency, especially for severe illness. However, there is strong evidence from our study (Table 1) that knowledge of the CEP, and to some extent the recognition of danger signs (Table 2), vary substantially according to a number of important social, geographic, and asset-related characteristics, although these are weaker in adjusted analysis because of the strong interrelationships. The fact that municipality, maternal literacy, and educational level lose significance in the multivariable analysis of factors associated with care seeking should not be interpreted as meaning that these are unimportant for policy.

Integrated Management

Awareness-raising interventions are consistent with the Integrated Management of Childhood Illness guidelines,³⁵ which recommend a 3-pronged approach. The component of "improving family and community practices through education of mothers, fathers, other child caretakers, and members of the community, with a focus on health seeking behaviour, compliance, care at home and on overall health promotion"³⁶ relates closely to the survey findings reported in this article. Supply-side results relating to the other 2 components, improving case management skills of health care staff and improving overall health systems, will be reported separately.

An expanded role for community health workers offers one means of raising awareness, especially in more remote social and geographical groups. In the poorer municipalities of Guatemala, working alongside contracted nongovernmental organizations, the community health worker (promotor de salud) has been a key component in the EdC. This service has historically focused mainly on prevention, but was also intended to contribute to case recognition and treatment. In this study, whether or not they were aware of this component of the role of EdC, parents sought advice and treatment from this source in very few cases. Furthermore, the few consultations resulted in almost no referrals to the health centers. This indicates that, in practice, EdC is not providing an effective or trusted community health worker–based service. This component of the Guatemalan health service is currently being reassessed. In doing so, it will be important to determine what arrangements can best fulfil the community health worker role.

Conclusions

The factors most strongly associated with care seeking were knowledge of CEP, and recognition of severity of the illness, whether this was the mother’s own perceptions, or knowledge of the WHO danger signs. Having someone to look after the family in an emergency was also important, and is a topic covered by the CEP. Greater distance (assessed by time on foot) from the health centers, which are located in the main town in each municipality, was a barrier. These results suggest that the CEP may well be effective, although the current study design does not allow a firm cause-and-effect association to be assumed. Awareness of the CEP, assessed by a simple question on recognition (and showing the plan), was reported by 70.1% in San Lorenzo, but reported by a little more than half (55.2%) of homes in Comitancillo, so it needs to be extended. Responsibility for promoting the CEP and increasing awareness of the need to seek early medical advice lies primarily with MSPAS, but could be enhanced by working with community leaders, and involving pharmacies and schools. Bilingual (Mam, Spanish) community health workers have a key role, not only in awareness raising, but also in being easily available to make initial assessments of sick children and supporting families to overcome barriers to early facility-based care seeking.

Underlying the importance of the CEP and knowledge of danger signs is the fact that these remain strongly associated with poorer socioeconomic conditions, and more remote geographical location. As a consequence, actions taken by MSPAS should be designed to respond to where need is greatest, but also be complemented by efforts to improve these underlying conditions, including women’s education and language skills, transport and road infrastructure, and financial constraints. A systems approach is helpful,³⁷ as it promotes coordinated action in respect of health services, community health workers, traditional providers (including pharmacies, shops, and healers), community development, education and literacy, and transport.